Episode #21: Anthropic's Latest Raise, Internal Turf Wars in Big Tech, IPO Alternatives In Venture

I think the venture business historically has been built on the IPOs are what create your legendary funds. Being in an Uber or a Twitter or a Google or a Facebook, that's how you create legendary funds. But what happens if those IPOs just don't happen? Or if they happen 20 years after formation? In a world where most venture funds have 10 year lives, what do you.

What do you do? It's the week of February 24th. Welcome to the Learning Corner, a Precursor Ventures podcast.

Welcome to the Learning Corner by Precursor. I'm Mia Farnam and I'm joined by our GP and Managing Partner Charles Hudson, where we will be talking through the articles and Goodreads that caught our team's attention over the past week. Topic number one Amidst the frenzy a month back with Deep Seat, According to Reuters, Anthropic is finalizing a $3.5 billion funding round that will bring its valuation to $61.5 billion, showing there is still eagerness from investors to continue funding the more expensive foundational models headquartered in the us. I never thought anything going on with Deep SEQ or probably any other competitors in this space would dry up any interest in OpenAI, anthropic, et cetera. But it's even more clear to me now that the carrot at the end of the stick of strategic positioning that is won from being part of these rounds is such an important part of this investment mindset. So the funding round is actually even larger than what Reuters reported last month by 1 billion. So curious to know where your thoughts are since we have a continuing thread about this on our team and in the podcast.

I've long felt like a lot of this big ticket large model stuff is a big poker game and you gotta have the chips to stay in the game. And if you don't have the chips, you can't invest in the next generation of your model. And if you can't invest in your next generation of your model, you're not competitive. And I think at some point this year or next year we will find out is there any residual value in a LLM provider who can't keep going? To me, it's enough level that at least one or two of the reasonably well funded people will fall behind and won't be able to continue to raise this money. I think it's interesting to see how this all plays out. The other thing that's crazy to me to think about is even if you'd invested in this company at a 5 or 10 billion valuation, which would have felt hard to imagine, you'd be up somewhere between 5 and 10x your money on paper. And I think right now, under the current thinking, the LLMs feel like the big winners in this category for a lot of people. And they're like, hey, hey, if these end up being the Google or the AWS of AI, this might be the best. And I also just think we're in a world where bigness begets bigness. You get to a certain size and if you get to a billion dollars, the probability that you'll get to $5 billion in this environment is actually reasonably high if you continue to execute due to the supply of capital. And look, if you're like, hey, I want to bet on an LLM provider that's likely to be one of the winners, provided that these high capital LLM models end up being long term valuable. The two best bets to make that are available are Anthropic and OpenAI. And anything after that is highly speculative.

It's crazy what you said before. From the last round to the current round, you could have been at 5x. You also could have probably made some liquidity if you wanted to, because everyone would be clamoring to come and take up whatever allocation is available, regardless of what happens in the news or what we see with deep seek, et cetera. I guess from my vantage point, I was starting to see in this current round, Anthropic starting to carve out maybe a more distinctive consumer approach with lower price points for computing power hybrid reasoning models, one of which they just released out this week, and even relatable benchmarks like the Pokemon tests that they did. But is it truly for differentiation or is it just for marketing? Kind of just wondering, are we watching actual competition for market share? Or if both Anthropic and OpenAI are essentially just running the same playbook, raising tons of unprecedented capital to secure computing resources and talent, while the business model itself is secondary. So are investors of both companies just funding genuine alternatives to one another, or are they just inadvertently creating an AI duopoly where the primary competition is more so for capital than customers?

Yeah, I think you took the words right out of my mouth. But I also think if I were a large customer who needed an LLM solution, I would want deepseek to be healthy, I would want Anthropic to be healthy, I'd want OpenAI to be healthy, I'd want Gemini to be healthy, and I'd want Llama to work for a couple reasons. One, I'd want the ability to price compare across those services, just like we see in Our portfolio companies, you know, it'd be a different world if AWS was the only option. But you've got gcp, you've got Azure, you've got for the really large people, the ability to self host. And I think actually that company competition is good for consumers, but it's also good for those companies. I'm sure their bottom lines would be better with less competition, but the presence of other people competing for that business forces them to push costs down or provide better service, or continue to invest in new products and services. So I think it would be customers who would benefit the least if this was a monopoly situation. Because then not only would you be stuck with one model, you'd be forced with the prices offered by those models or the model providers. So I think customers will continue, I think, to support multiple model providers because it's in their best interest to have ch.

I mean, we see it on our team internally when we're using different tools. You and I both, I feel like subscribe to a lot of the different models and pay the monthly price to get the pro version of it all. And I think a lot of that is to do with, we like to compare and see what the differences are. My question is at what price point would somebody not be as curious to leverage it? And this is truly more so from like a D2C angle, not from an enterprise level at all.

I'll say this, I remember when Spotify came out and my friend was like, can you imagine that you could have basically all of the world's music in your pocket for like 10 bucks a month? That seems like a criminally cheap price. And if I told you you could have the world's smartest supercomputer in your pocket for 20 bucks, I think most people would say that's a criminally cheap price. And the thing that's been eye opening to me is I know a lot of students who pay for ChatGPT. You know, students are cheap. But I think if a student tells you that ChatGPT is worth 20 bucks a month, it's really worth at least 20 bucks a month to them. And I think the more interesting thing is as ChatGPT does more as perplexity pushes the envelope, like how many of these things are people going to pay for? As I explained to our cfo, I actually really need to pay for all of these things for testing and lear purposes. I spend hundreds of dollars a month on AI tooling. That is not normal. I don't expect most people to do that either. In a business or a professional context. I think most people will end up paying for one of these things somewhere between zero and one of these products. And I will say this, Mia, I scoffed at paying for ChatGPT Pro and then I said I should just pay for it and see if it's different. It is pretty sweet. I'm like, wow, for 200 bucks I get to experiment with an agent. I get essentially unlimited use of deep research. I get all of these other models. Is it worth 200 bucks? The way I've been using it, it's actually worth more than that to me.

The interesting part is now also we're seeing folks. Not you and I. Unfortunately we're not that young, but folks who are growing up with these tools. They probably won't know an alternative without them. They've only grown up in a world where access to those types of models is priced at 200amonth, and so they can't imagine what it's like. So that's normal. It'll be really interesting to see how it all flows generationally too, even from a macro perspective. Our second topic was by Aaron Wu at the Information titled Google's AI Efforts Marred by Turf Disputes, the article details how Google's AI development is being hindered not by the capacity and ability to build, but rather internal organizational tensions. Aaron walked through what non Google folks didn't see when NotebookLM was launched and hailed as a success, according to the article, apparently there were clashes between the Google Labs experimental unit and folks on the Workspace team, all wanting to draw specific parameters around what turf they own and what innovation capabilities should be brought in by outsiders. The tension expands to to other areas DeepMind and Google Cloud sharing conflicting priorities Sundar having to scale back Pixel's AI assistant to not internally compete with the Gemini Assistant. Despite reorganization efforts to consolidate AI initiatives under DeepMind, these conflicts persist. So from the vantage point of this article, we are really seeing how the frenzy and opportunity around AI innovation has created such tension internally. However, it's clear that dynamics like these arise simply when a company becomes massive. And there was a great quote by Carl Bass, who's the former Autodesk CEO and known former special advisor to Larry Page, who was quoted saying, big companies get in their own way. Small companies have a much more defined path to getting stuff done. So I was curious what you thought here.

I wish I could like cite the person who said this, but I don't think I have their permission, so I won't. I talked to someone who's looked at a couple of these Coding co pilots like Cursor. And I was like, how does Cursor be GitHub Co pilot, like GitHub Copilot has all the developers because of GitHub. And this was before this article came out. I don't think this person had any proprietary insight into this notebook. He said, we're betting that in some of these cases, the turf wars and political battles between product owners in these companies will inhibit their ability to ship product and that these companies will succeed in spite of all of the distribution advantages that these big companies have. They will succeed. And I thought about one really specific example that I lived through. I worked at Google at a time when Google dropped Drive was known as Platypus. And Google Drive was under development and there was a lot of time and effort put into it and Dropbox came out and everyone's like, we're going to like crush these. Like, there's no way this little startup company. And the thing you realize at a big company is not only do you have to fight all these political battles, you also have to deal with the problems of scale, which is if you're Dropbox, you can say, hey, this is the thing. It kind of works. It only works on these computers. It only works in these. You can launch a constrained product when you're a Google or a Microsoft or a faang company and you roll out. It kind of has to work for a lot of different edge cases and use cases and there's a level of polish and finish that people expect even on V1. And it's got to be multilingual and it's got to work. It's got to comply with these data requirements. You end up having to do so much work and then when you launch it, you can't quietly launch it in Tinker because you launch it and everybody notices. And the question is, how do you know if your Snapchat? And you're like, we have stories, but Facebook's going to copy stories, but it's not going to kill Snapchat versus the little featurettes that sometimes get crushed when a big company announces a new product capability. And my general rule of thumb is you never want to. As my friend used to say, you never want to compete with a shark in the ocean. You don't want to compete with a big company in their core domain. I think until very recently, it was nearly impossible to compete with Google and Core Internet search. They effectively beat back everybody who attempted to do that. And Perplexity and chatgpt are probably the first real threats they faced. And so I think we will end up seeing a really mixed bag here. There'll be some great Harvard Business School case studies, and maybe it's cursor, maybe it's replit, maybe it's, I don't know, pick your favorite AI powered tool that's competing with an incumbent that has a distribution advantage. There'll be some story about how internal politics or some other miscalculation by a big company created space for somebody to really get a foothold. And there'll be other examples where the distribution advantage of the incumbent snuffed out the entire opportunity almost immediately. So. So I'm very curious to see how this plays out. And I think if you really know the inside of these companies and the personalities of the leaders, you might be slightly more attuned to picking out the ones where it's worth going toe to toe.

But you have to wonder. Google must have reasons why they invested in anthropic. Even though they have tons of money going into their own in house AI tools, it's almost saying some of the things that maybe they can't acknowledge themselves about how quickly they need to scale a product in a way that works at the scale that people want it. So taking bets outside their business probably makes sen.

I think it was interesting. I was just reading an article. I don't remember where I saw it. I might have saw it on Blue Sky. It was about Amazon shutting down their Steam competitor after putting a quarter of a billion dollars into it. And they were like, it sort of never occurred to us that we wouldn't win. And the person who wrote in the article there was a big part was like, well, we relied so much in distribution that we didn't really do all the things you're supposed to do. Make sure this was a product that people wanted and that we understood why they wanted it. And I do think sometimes big companies overestimate the value of distribution and can skip really important steps about customer discovery and like building something that people want something to just put something out there when you know that it's going to get distribution.

Taking an entirely different perspective on it. I mean, their core issue just seems to be balancing innovation with protecting existing revenue streams. So they're essentially just neutering their own AI products to avoid cannibalizing on the current businesses. Is this strategy ultimately just self defeating in a really quickly evolving market? Or is there room or possibility that this cautious approach could actually be more prudent than the move fast philosophy that I feel like is just sweeping across AI Right now I think one of.

The hardest things to do when you're running a business that's working is you can totally see the future. And I think lots of businesses think, oh, we'll let it continue to erode and we'll try to cut over just at the moment where the new thing is working and it never works because you can't predict the rate of erosion. That old saying about like, it's hard to get a man to believe one thing when his paycheck believes in something else being true. And I think it's really hard for big companies when you're like we have just this massive business doing thing X. And if you look at the history of most tech innovation, that's how it's gone. You know, when Google bought rightly and started doing docs and spreadsheets, those products were not as good as Excel and Word. They had one real advantage. They were much better at real time collaboration. And there are a bunch of things that Google did in the cloud that were inferior to desktop products until everything moved to the cloud. And then there are a bunch of things in social that like meta understood much better than Google did. And it seems to me that there are some things in this AI world that the upstarts seem to understand or be able to grasp and implement faster than people who have to your point, a business to defend. And it's really hard as a public company to say we're going to put the entire franchise at risk to chase something that's still kind of unproven.

Moving on to the last topic. I don't have a specific article on this just because it's a thread that everyone has been talking about honestly since probably Q2, 2024, but it's focused on the lack of IPOs as we kind of barrel through Q1 of 2025. I mean 2025 was seen as the redeeming year for IPOs for most venture capitalists after such a drought over the last few years. So we're almost through February. Somehow an IPO activity and announcement has not turned out in the way that most individuals were hoping for. I know on beh the VC land for sure. So there are still a lot of big names expected to go public in 2025. So you have like stripes, databricks, klarna, all folks that people would love to see them announce. However, I think there are broader questions around IPOs as the main route for venture backed companies to return capital back to investors. And I know that you have some thoughts specifically around the area So I wanted to kind of dive deeper with that. With you.

Yeah. I would encourage people to read Aaron Griffith's article. I thought it was really good in framing this. I think it's going to be interesting. A lot of people said the second that we have the administration, all of the animal spirits will be unleashed and there will be pent up deal demand. And I'm like, I don't think that's true. I don't think antitrust and regulatory was the only reason deals weren't getting done. I certainly believe that that was a contributor in some cases.